How to measure “non-public” opinions? Examining sensitivity bias of conventional surveys in an authoritarian context

It is always difficult to understand what people think about sensitive political and social subjects in societies under authoritarianism. The use of public opinion surveys via phones/interviews is a rigorous method in free societies to extract attitudes and preferences from a population. However, in suppressed societies, many opinions about sensitive questions are dangerous to get "public" and should be kept "non-public." So, the challenge remains how "non-public opinions" can be measured and revealed in such societies so that people express their genuine attitudes without fear of consequences from the regime.
The deficit of and demand for understanding non-public opinions in influential authoritarian countries stimulated institutions and researchers to utilize the conventional sampling methods without considering their limitations and misleading results. Knowing this demand, many authoritarian regimes either themselves invest in conventional polling as a propaganda tool or be thankful for international pollsters' results, trying to buy legitimacy by selling indicators such as high electoral turnout and high approval rate of their leaders.
This study examines the results of three conventional surveys conducted in Iran by reputable World Values Surveys in 2020 and Gallup in 2021, and Canada-based IranPoll in 2021. The paper will present and compare the results of large-scale, well-designed, and well-adjusted online anonymous surveys by GAMAAN among Iranian residents in 2021-2022, replicating the questions from the aforementioned surveys. The findings show a high discrepancy of about 50 percentage points for different sensitive questions while a high congruency for various non-sensitive questions.
Trying to understand whose public opinions are reflected by the conventional surveys, we found that their results are substantially similar to the attitudes of the sub-sample of online surveys that support the regime. The paper's findings have important implications for surveying research in authoritarian contexts.